Introduction to Game Theory - eBooks and textbooks from bookboon.com

In contrast to pure strategies, in mixed strategies a player does not decide on one pure strategy, but plays several pure strategies, each of which has specific probability.. A classical[r]

Trang 1

Download free books at

Trang 2

Christian Julmi

Introduction to Game Theory

Download free eBooks at bookboon.com

Trang 3

Introduction to Game Theory

ISBN 978-87-403-0280-6

Trang 4

Contents

Click on the ad to read more

www.sylvania.com

We do not reinvent the wheel we reinvent light.

Fascinating lighting offers an infinite spectrum of possibilities: Innovative technologies and new markets provide both opportunities and challenges

An environment in which your expertise is in high demand Enjoy the supportive working atmosphere within our global group and benefit from international career paths Implement sustainable ideas in close cooperation with other specialists and contribute to influencing our future Come and join us in reinventing light every day.

Light is OSRAM

Trang 5

360°

Discover the truth at www.deloitte.ca/careers

Trang 6

We will turn your CV into

an opportunity of a lifetime

Do you like cars? Would you like to be a part of a successful brand?

We will appreciate and reward both your enthusiasm and talent.

Send us your CV You will be surprised where it can take you.

Send us your CV on www.employerforlife.com

Trang 7

1 Foreword

This book has set itself the task of providing an overview of the field of game theory The focus here is above all on imparting a fundamental understanding of the mechanisms and solution approaches of game theory to readers without prior knowledge in a short time Because game theory is in the first place a mathematic discipline with very high formal demands, the book does not claim to be complete Often, the solution concepts of game theory are mathematically very complex and impenetrable for outsiders However, as long we remain on the surface, some principles can be explained plausibly with relatively simple means For this reason the book is eminently suitable in particular as introductory reading, so that the interested reader can create a solid basis, which can then be intensified through advanced literature

What are the advantages of reading this book? I believe that through the fundamental understanding of game theory concepts, the solution approaches that are introduced can enlighten in nearly all areas of life – after all, along with economics, it is not for nothing that game theory is applied in a huge number

of disciplines, from sociology through politics and law to biology

With this in mind I hope you have a lot of fun reading this book and thinking!

Trang 8

2 Introduction

2.1 Aim and task of game theory

Game theory is a mathematical branch of economic theory and analyses decision situations that have the character of games (e.g auctions, chess, poker) and that go far beyond economics in their application The significance of game theory can also be seen in the award of the Nobel prize in 1994 to the game theoreticians John Forbes Nash, John Harsanyi and Reinhard Selten

Decision situations usually consist of several players who have to decide between various strategies, each

of which influences their utility or the payoffs of the game The primary aim here is not to defeat fellow players but to maximise the player’s own (expected) payoff Games are not necessarily modelled so that the gains of one player result from the losses of the opponent (or opponents) These types of games are simply a special case and are referred to as zero-sum games

Game theory is therefore concerned with analysing all the framework conditions of a game (insofar

as they are known) and, taking account of all possible strategies, with identifying those strategies that optimise one’s own utility or one’s own payoff The decisive point in game theory is that it is not sufficient to consider your own strategies A player must also anticipate which strategies are optimal for the opponent, because his choice has a direct effect on one’s own payoff There is therefore reciprocal influencing of the players In the ideal case there are equilibriums in games, which, roughly speaking, means that the optimal strategies of players ‘are in harmony with one another’ and are ‘stable’ in their direct environment This obviously does not apply to zero-sum games such as ‘rock, paper, scissors’, in which no constellation of strategies is optimal for all players

In classical game theory it is assumed that all players act rationally and egoistically According to this, each player wants to maximise his (expected) benefit The final chapter shows that this does not always conform to reality

2.2 Applications of game theory

There is a series of applications of game theory in different areas Game theory is above all interesting where the framework conditions can be easily modelled as a game, that is, in which strategies and payoffs can be identified and there exists a clear dependency of the payoffs of the different players on the selected strategies

Trang 9

of evolutionary game theory The latter models how successful modes of behaviour assert themselves in nature through selection mechanisms, and less successful ones disappear.

A classical example of game theory modelling (and unfortunately not applied) in economics is the auction

of UMTS licences in Germany in 2000 The licences were distributed between six bidders for a total of

DM 100 billion – a sum that dramatically exceeded expectations The high price also signalled the great expectations regarding the economic importance of the UMTS standards, but could have turned out much less, because in the end the six bidders bid each other up to induce other bidders to drop out However, because in the end no one dropped out, the high price had to be paid without an additional licence The book by Stefan Niemeier Die deutsche UMTS-Auktion Eine spieltheoretische Analyse published in 2002 shows, for example that from a game theory aspect the result is not always based on rational decisions, and that, given a suitable game theory analysis, some bidders could have saved money

2.3 An example: the prisoner’s dilemma

Probably the most famous game theory problem is the prisoner’s dilemma, which will be introduced briefly here, and which provides an initial impression of how games can be modelled Essential terms will also be introduced that are important for reading the following chapters

Two criminals are arrested They are suspected of having robbed a bank Because there is very little evidence, the two can only be sentenced to a year’s imprisonment on the basis of what evidence there is For this reason, the two are questioned separately, with the aim of getting them to confess to the crime through incentives, and because of the uncertainty regarding what the other is saying A deal is offered

to each of them: if they confess, they will be freed – but only if the other prisoner does not confess; in this case he will go down for 10 years If they both confess, they will each go to prison for five years

Trang 10

The terms introduced up to now enable some statements to be made on the game theory modelling of this game The two criminals are two players, each of whom has two strategies available: to confess or not to confess Their payoff corresponds in this case to the years that they will have to spend in prison, whereby here, of course, the aim is not to maximise the payoff but to minimise it The payoff depends not only on a prisoner’s own strategy but also on the strategy of the other prisoner It is also important that the two criminals make their decisions simultaneously and that each of them is unaware of the other’s decision In addition, this information is known to both players Games like this are known in game theory as simultaneous games under complete information Simultaneous games are also referred

to as games in normal form, while sequential games – in other words, games in which ‘play’ takes place sequentially – are known as games in extensive form Because two persons play the game, it is a 2-person game or a 2-person normal game

With this information, the following model can be set up using game theory:

1 Both prisoners confess (top left field)

2 Prisoner 1 confesses, prisoner 2 does not confess (top right field)

3 Prisoner 1 does not confess, prisoner 2 confesses (bottom left field)

4 Neither prisoner confesses (bottom right field)

The two numbers in the four fields correspond to the payoffs of the two prisoners The payoffs in the bottom left accrue to prisoner 1 in the respective constellations, while the payoffs in the top right are for prisoner 2

Trang 11

So much for the notation But what is it about this game that has enabled it to become so famous? The response is found in the paradoxical result that this game entails, namely that both confess and go to prison for five years, although if they had just said nothing, they would each have been sentenced to only one year’s imprisonment

We arrive at this result if we consider a prisoner’s strategies more exactly from the aspect of the other prisoner Let us assume that I am prisoner 1 I then consider my best response for each of the other prisoner’s two strategies If prisoner 2 confesses, I will confess as well, because in this case I will only have to go to prison for five years, instead of 10 years if I do not confess In contrast, if I assume that prisoner 2 will not confess, I will confess myself, because I will then be released, which I naturally prefer

to going to prison for one year, if I confess as well This means I always choose the ‘confess’ strategy, completely regardless of which strategy the other prisoner chooses Because the same case applies to the other prisoner, he will also confess, which leads to the paradoxical result described above

This case can, of course, be regarded as a construction that is relevant only in theory However, this can

be countered by saying that life is full of prisoner’s dilemma, namely whenever two (or more) parties

do not move from their positions because they are afraid of being the only party to make concessions while the other parties do not move (for example, between management and union representatives)

2.4 Game theory terms

2.4.1 Preferences

Preference relations are extremely important in game theory They state which alternatives a player prefers to other alternatives, and to which alternatives a player is indifferent If a player prefers strategy (A) to strategy (B), we write A > B ; if he is indifferent with regard to both strategies we write A ~ B

Let us assume that a player has the choice of travelling by car (A), bus (B) or tram (C) The following is

to apply with regard to his preferences:

1 The player prefers to travel by car rather than by bus: A > B

(“The player prefers A to B”)

2 It is all the same to him whether he travels by bus or tram: B ~ C

(“The player is indifferent with regard to B and C”)

Because of transitivity, A > C then follows from (1) and (2)

Trang 12

S is described in this case as a strategy pair as well

S1 (S2) may also stand for a set of strategies of player 1 (player 2) to choose from, for example:

S1 = (confess, do not confess)

as a

e s

al na or o

eal responsibili�

�e Graduate Programme for Engineers and Geoscientists

as a

e s

al na or o

Month 16

I was a construction

supervisor in the North Sea advising and helping foremen solve problems

I was a

he s

Real work International opportunities

�ree work placements

al Internationa

or

�ree wo al na or o

I wanted real responsibili�

I joined MITAS because

www.discovermitas.com

Trang 13

and for player 2:

S2 = (confess, do not confess)

and for the whole game

S = (S1, S2) = ((confess, do not confess), (confess, do not confess))

If the actions of the players in a game consist of the decision for one of the available strategies, we speak

of a pure strategy The strategies of the game itself are also referred to as pure strategies In contrast, if several pure strategies of a player are each played with a certain probability, we speak of a mixed strategy

A classical example of a game in which the player pursues a mixed strategy is game ‘rock, scissors, paper’.2.4.3 Payoffs

Payoff A is used below to designate what is ‘paid out’ to a player on a given constellation of strategies Because the payoff depends not only on a player’s own strategy, but also on the strategies of all other players, payoff A is a function over strategy of all players

For the prisoner’s dilemma the payoff for player 1 would then be:

The payoff is therefore dependent not only on a player’s own strategy, but also on the strategy of all players

Trang 14

The following sections provide an overview of the different types of strategies and equilibriums in a two-person simultaneous game Although at first only simultaneous games between two persons will be discussed – because they can be represented in a two-dimensional matrix – multi-person games (n-person simultaneous games) for which the same principles and mechanisms apply are also possible, as will be shown in conclusion in this chapter by means of a three-person simultaneous game.

3.2 Strategies

3.2.1 The maximin strategy

The maximin strategy corresponds to that strategy of a player with which he still achieves the best payoff

in the most unfavourable case Its objective is therefore damage limitation

The maximin strategy can be determined in a matrix with any number of strategies n for player 1 and

m for player 2 in two steps:

1 First off all, the smallest possible own payoff (min A) is selected for each own strategy taking account of all possible strategies of the opponent If this occurs more than once, these are to be selected accordingly

2 Following this, the largest (max min A) of these smalle st payoffs is selected The

corresponding strategy is called the maximin strategy of the corresponding player Several maximin strategies can exist for one player

The maximin strategy of player 1 is designated MS1, the corresponding maximin strategy of player 2

as MS2

Trang 15

PLQ$=6

Trang 16

PLQ$6=

The maximin strategy for player 2 is therefore strategy X2 The maximin strategy is the strategy with the least risk of a small payoff, without making assumptions about the preferences of the opponent (or opponents)

3.2.2 Dominant strategy

A strategy is designated as a dominant strategy if it holds for every other strategy that the latter do not put the player in a better position, and put him in a worse position in at least one case A dominant strategy

is thus ‘resistant’ to any possible change of strategy by the opponent, and is selected in each instance

We can find an example of a dominant strategy in the example of the prisoner’s dilemma shown above

In this game, the dominant strategy for the prisoner is to confess, because in each instance this strategy puts him in a better position than the alternative strategy of not confessing

A modification of the prisoner’s dilemma also provides a good illustration of the principle of the dominant strategy In this modified version, both prisoners will definitely go to prison for 1 year As soon as one

of the two confesses to the crime, both must go to prison for 5 years This situation can be mapped in the following matrix:

S1 = (Do not confess) is the dominant strategy for prisoner 1, because on a change of the strategy to

S1 = (Confess) he is by no means in a worse position – it is of no concern if prisoner 2 chooses strategy

S2 = (Confess) – and in at least one case – with S2 = (Do not confess) – he is better off

Trang 17

The following applies therefore

S1 = (Do not confess) is the dominant strategy for player 1

S2 = (Do not confess) is the dominant strategy for player 2

If a dominant strategy for a player exists in a game, this player will always select the dominant strategy.3.2.3 Dominated strategy

A dominated strategy has the characteristic for a player in a game that there is another strategy in this game that in each instance – that is, with every possible strategy of the opponent – is not worse, and

is really better in at least one case A dominated strategy can be removed from the matrix for further analysis, because in no case does it bring an advantage for the player in comparison with the strategy that dominates it

The following example is intended to illustrate this situation:

It is easy to understand that player 1 prefers strategy Y1 to strategy X1, because it either places him at

an advantage (if player 2 chooses X2 or Y2) or not in a worse position (if player 2 chooses Z2) From the point of view of player 1 therefore:

applies correspondingly for the preferences of player 1 with regard to his strategies

Trang 18

Player 1 will therefore select Y1 in preference to X1 We can also say strategy X1 is dominated by strategy

Y1 A statement cannot be made regarding Y1 and Z1 Which strategy player 1 prefers here depends on the strategy that player 2 selects

As player 1 will never play X1 because he prefers Y1, strategy X1 can be deleted from the game:

STUDY AT A TOP RANKED INTERNATIONAL BUSINESS SCHOOL

Reach your full potential at the Stockholm School of Economics,

in one of the most innovative cities in the world The School

is ranked by the Financial Times as the number one business school in the Nordic and Baltic countries

Visit us at www.hhs.se

Swed

Stockholm

no.1

nine years

in a row

Trang 20

Let us take another look at the prisoner’s dilemma:

BR2(Do not confess) = (Confess)

If we start from a strategy and then determine the best responses – that is, the best response to a strategy, the best response to this best response, etc – this process can be continued until the best response for

a strategy pair is in each case the best response to itself as well Because then an equilibrium has been found in which it is not worthwhile for any player to deviate unilaterally from this equilibrium We will come across this situation again soon with the Nash equilibrium

Trang 21

The fact that this process does not always end in a state of equilibrium is shown by the following example

in which, starting from a specific strategy pair S = (Y1, X2), a ‘best response circle’ follows that does not lead to a stable equilibrium:

Trang 22

3.3.2 Nash equilibrium

The Nash equilibrium is one of the central solution concepts of game theory It was developed in 1950

by John Nash and is relevant far beyond simultaneous games in nearly all forms of games A Nash equilibrium is found when with a given strategy, it is not worthwhile for a player to be the only one to change his strategy Therefore, in the Nash equilibrium no player has an incentive to depart from this equilibrium unilaterally The expression “unilateral deviation is not worthwhile” can be used as a rule

of thumb for the Nash equilibrium

A Nash equilibrium does not make any statement as to whether the players can position themselves better if at least two players deviate from their strategy simultaneously

Our prisoner’s dilemma serves once again as an example:

as long) Only if both were to change their strategy would an incentive to change arise

For Nash equilibriums we use the notation as shown in this example

Trang 23

But how can the Nash equilibrium be determined practically and methodically in a simultaneous game

of two persons with any number of strategies? The easiest way is first of all to go through all possible strategies of player 2 for player 1, and in each case to mark the best payoff for player 1 for the various strategies (where there are several equally large best payoffs these are to be marked accordingly) This process is then applied vice versa for player 2 In this way, the own best responses (or rather, the payoffs belonging to them) are marked for all possible strategies of the opponent The Nash equilibriums are then all those fields in which both payoff sizes are marked

This method will be shown by means of the following example:

Trang 24

In this way, 6 values are now marked, whereby the two payoffs are marked in the fields (Y1, Y2) and (Z1,

Z2) These fields therefore each mark a Nash equilibrium, so that the game contains the following two Nash equilibriums:

S* = (Y1, Y2), S1* = Y1, S2* = Y2

S** = (Z1, Z2), S1** = Z1, S2** = Z2

The six steps shown here for determining the Nash equilibriums are shown once again in the following illustration:

Trang 25

a single best response to the strategies of his opponent This means that there cannot be several strict equilibriums in a game.

The following shows the existence of a strict Nash equilibrium:

In general, S* = (X1, X2) is held to be a strict Nash equilibrium if not only X1 but also X2 is a dominant strategy Strict equilibriums are always Nash equilibriums as well

3.3.4 Nash equilibriums and best responses

Nash equilibriums and best responses are directly connected The following example serves to make this clear:

Trang 26

3.4 Equilibriums in mixed strategies

3.4.1 Mixed strategies and expected payoffs

In contrast to pure strategies, in mixed strategies a player does not decide on one (pure) strategy, but plays several pure strategies, each of which has specific probability

A classical example of a game in mixed strategies is tossing a coin in which there is a 50% chance of heads or tails being on top Assume that a player tosses a coin If it turns up tails he receives $2, if heads,

he receives nothing The coin receives nothing in this game The payoff matrix then looks like this:

Trang 27

SCoin = (½, ½)

With this notation the strategies are no longer given but instead the probability with which the strategies are played Anyone who is unable to imagine that a coin can pursue a strategy can imagine instead a player who decides on the strategies X2 and Y2 by tossing a coin in each case

How high is the player’s expected payoff now? Because the player has a 50% chance of receiving $0 (heads) and a 50% of receiving $2 (tails), his expected payoff is:

½ · $0 + ½ · $2 = $1

For the expected payoff we write:

E(Aplayer(toss coin, (½, ½))) = ½ · 0 + ½ · 2 = 1

(“The expected payoff of the player when he tosses the coin is 1”)

A slightly more complex example should make this situation clearer:

“The perfect start

of a successful, international career.”

Trang 28

(“The expected payoff of player 1 if he plays X1 and player 2 plays S2 = (¼, ¾) is 11/4”)

Let player 1 now decide to play a mixed strategy S1 = (½, ½) as well Each strategy combination then occurs with a specific probability:

Trang 29

3.4.2 Mixed Nash equilibriums

Mixed Nash equilibriums are Nash equilibriums that consist of mixed strategies This can be illustrated very well with the ‘rock, scissors, paper’ game, in which rock beats scissors, scissors beats paper and paper beats rock The winner receives a payoff of $1, which the loser has to pay If they both chose the same strategy, no one gets anything The payoff matrix then looks like this:

is there no incentive for any player to change his strategy

The strategy

S* = ((1/3, 1/3, 1/3), (1/3, 1/3, 1/3))

is accordingly a Nash equilibrium that consists of mixed strategies and in which no player (as shown) has an incentive to deviate from this strategy

In principle, every game has a Nash equilibrium If there is no Nash equilibrium in pure strategies, there

is at least one Nash equilibrium in mixed strategies However, a game can also have Nash equilibriums

in pure and mixed strategies

Trang 30

But how is a mixed Nash equilibrium determined? For this purpose, the following simple game is

to be considered in which player 1 plays strategy X1 with the probability p and strategy Y1 with the corresponding probability (1-p), and player 2 plays the strategy X2 with the probability q and strategy

Y2 with the corresponding probability (1-q):

Y1(1-p)

This game has no Nash equilibrium in pure strategies However, because every game has at least one Nash equilibrium, there exists at least one mixed Nash equilibrium

89,000 km

In the past four years we have drilled

That’s more than twice around the world.

careers.slb.com

What will you be?

Who are we?

We are the world’s largest oilfield services company 1 Working globally—often in remote and challenging locations—

we invent, design, engineer, and apply technology to help our customers find and produce oil and gas safely.

Who are we looking for?

Every year, we need thousands of graduates to begin dynamic careers in the following domains:

n Engineering, Research and Operations

n Geoscience and Petrotechnical

n Commercial and Business

Trang 31

To determine a mixed Nash equilibrium considerations must be made regarding the probabilities with which a player has to play his strategies so that his opponent is indifferent with regard to these strategies Looking back at the ‘rock, scissors, paper’ game, this means that I have to select my strategy in such a way that my opponent does not prefer a specific strategy with which he can outsmart me – which would then give me an incentive to change the strategy

Coming back to the example: let player 1 now consider which strategies he has to play so that player 2

is indifferent with regard to his strategies, i.e the expected payoffs for player 2 must be equal for X2 and

Y2 This results in the following equation:

First of all, we assume that player 1 plays rock with probability p, scissors with probability q and paper with probability (1-p-q):

Trang 32

Y1: Scissor (q)

Z1: Paper (1-p-q)

(1) E(A2((p, q, (1-p-q)), Rock)) = E(A2((p, q, (1-p-q)), Scissor))

(2) E(A2((p, q, (1-p-q)), Scissor)) = E(A2((p, q, (1-p-q)), Paper))

(3) E(A2((p, q, (1-p-q)), Rpck)) = E(A2((p, q, (1-p-q)), Paper))

These three equations have to be fulfilled in the mixed Nash equilibrium We first solve equation (1):

Trang 33

American online

LIGS University

▶ enroll by September 30th, 2014 and

▶ save up to 16% on the tuition!

▶ pay in 10 installments / 2 years

▶ Interactive Online education

▶ visit www.ligsuniversity.com to

find out more!

is currently enrolling in the

Interactive Online BBA, MBA, MSc,

Note: LIGS University is not accredited by any

nationally recognized accrediting agency listed

by the US Secretary of Education

More info here

Z1: Paper (1-p-q)

(1) E(A2((p, q, (1-p-q)), Rock)) = E(A2((p, q, (1-p-q)), Scissor))

(2) E(A2((p, q, (1-p-q)), Scissor))... E(A2((p, q, (1-p-q)), Paper))

(3) E(A2((p, q, (1-p-q)), Rpck)) = E(A2((p, q, (1-p-q)), Paper))

These three equations have to be fulfilled in... X1 can be deleted from the game:

Download free eBooks at bookboon.com< /small>

STUDY AT A TOP RANKED INTERNATIONAL BUSINESS

Định dạng
Số trang	67
Dung lượng	3,46 MB